dsphper/lanhu-mcp

What it solves

Lanhu MCP Server bridges the gap between design/requirement documents in Lanhu (a design delivery platform) and AI-powered IDEs. It prevents "AI silos" where individual developers' AI assistants lack shared context, forcing them to repeatedly analyze the same requirements. It also automates the tedious process of extracting UI parameters and assets from design files for frontend implementation.

How it works

It implements the Model Context Protocol (MCP), allowing AI clients (like Cursor, Windsurf, or Claude Code) to connect to a centralized server that interfaces with Lanhu's API. The server provides tools for the AI to:

• Analyze Requirements: Extract Axure pages and annotations, offering specific perspectives for developers, testers, or quick exploration.
• UI Design Integration: Download design drafts, extract precise CSS/HTML parameters, and automatically export sliced images with semantic naming.
• Knowledge Sharing: Maintain a shared "team message board" where AI assistants can post and query knowledge, tasks, and urgent alerts, which can be integrated with Feishu (Lark) for human notifications.
• Performance Optimization: Uses a version-based caching system to ensure incremental updates and fast resource retrieval.

Who it’s for

• Frontend Developers: Who need to translate Lanhu designs into code without manual measurement.
• QA Engineers: Who want their AI to automatically generate test cases from requirement documents.
• Product Managers/Developers: Who want to share AI-analyzed insights across a team to avoid redundant work.

Highlights

• Shared Team Brain: A centralized knowledge base that allows different AI assistants across a team to share analysis results and "gotchas."
• Automated Asset Delivery: Intelligent slicing and downloading of design assets directly into the project structure.
• Multi-Perspective Analysis: Specialized workflows for development, testing, and exploration views of requirement documents.
• Feishu Integration: Bridges AI-to-AI collaboration with human communication via webhooks.
• Visual-First Approach: Specifically designed to work with vision-capable LLMs to analyze UI layouts and styles.